KR102387492B1 - Method for facial authentication of a wearer of a watch - Google Patents

Abstract

The present invention relates to a method for face recognition of a wearer of a watch (1), comprising the steps of:
- initiation (10) of an authentication process comprising the substep of detection (11) of at least one triggering movement/gesture performed by the wearer;
- capture ( 13 ) of at least one sequence of images of the wearer's face pivoting from one direction to another in front of the optical sensor ( 3 );
- obtaining (17) surface geometric data of the face associated with each image of said at least one sequence of images;
- generating ( 18 ) a three-dimensional model of the wearer's face from said captured at least one sequence of images and from said obtained surface geometric data;
- determination ( 19 ) of an identification index generated on the basis of the identification data for a plurality of features characteristic of the face of the wearer of the watch ( 1 ) detected on the basis of a three-dimensional model; and
- if the identification index is greater than the reference identification index, include the identification 21 of the wearer.

Description

METHOD FOR FACIAL AUTHENTICATION OF A WEARER OF A WATCH

The present invention relates to a method for facial recognition of a wearer of a watch.

The invention also relates to a computer program as well as a watch implementing this method.

A watch includes a set of functions that can be used by the wearer. These functions may relate to time or date adjustments as well as the use of confidential or private data useful for accessing services that are unique and personalized to the wearer. This data may be encrypted or authenticated by secure servers such as, for example, keys, certificates, authentication codes, passwords and personal codes capable of performing the watch's secure connection to the company's private network, and secure servers such as bank servers. A secure messaging service for sending and receiving emails and/or signed emails. As a result, it should be understood that it is important to be able to ensure access to the use of these watch functions.

To this end, authentication methods are known from the prior art and are designed to ensure access to the use of these functions of the watch by entering a secret code of the PIN type.

However, these authentication methods are not entirely satisfactory as the watch wearer entering the secret code can be observed relatively easily by a third party. This can be an embarrassing situation, as the mechanism also has to be memorized by the person wearing the secret code.

Accordingly, it is understood that there is a need for an alternative solution that does not particularly have the drawbacks of the prior art.

It is also an object of the present invention to propose a method for face recognition of a wearer of a watch, which is easy to implement, which offers robust and robust authentication of the wearer.

For this purpose, the present invention relates to a method for facial recognition of a wearer of a watch, comprising the steps of:

- initiation of an authentication process comprising a substep of detection of at least one triggering movement/gesture performed by the wearer;

- capture of at least one sequence of images of the wearer's face pivoting from one direction to another in front of the optical sensor;

- acquisition of surface geometric data of the face associated with each image of said at least one sequence of images;

- generation of a three-dimensional model of the wearer's face from the captured at least one sequence of images and from the obtained geometric data;

- determination of an identification index generated on the basis of the identification data for a plurality of features characteristic of the wearer's face of the watch detected on the basis of the three-dimensional model; and

- if the identification index is greater than the reference identification index, include the identification of the wearer;

In other embodiments:

- the step of initiating comprises a substep of determining that the wearer's face located opposite the optical sensor belongs to the living body;

- the substep of acquiring a plurality of images of all parts forming the wearer's face when the face pivots from one direction to another in front of the optical sensor;

- the sub-step of acquiring comprises a capturing step of capturing at least two images for each same part of the face having two different light exposures, one exposure for visible light and one exposure for infrared light ;

- The sub-steps to be obtained are:

■ A watch in a stable or substantially stable position;

■ the wearer's face optimally positioned with respect to the optical sensor for inclusion in each image of at least one sequence of images;

visual and/or sound guiding to the wearer of the watch with the goal of maintaining

- the acquiring step is performed simultaneously or substantially simultaneously with the capturing step;

- the determining step comprises a sub-step of processing wherein identification data originating from the three-dimensional model is compared with identification data pre-recorded against a reference three-dimensional model to determine correspondence between the identification data;

- said at least one triggering movement/gesture is performed by the part of the wearer's body comprising the watch during positioning of the wearer's face opposite the optical sensor of the watch.

The invention also relates to a particularly smart watch embodying this method, said watch comprising a processing unit, an optical sensor, at least one light source and behavioral and/or physiological sensors, said processing unit to the optical sensor; connected to said at least one light source and to behavioral and/or physiological sensors.

The invention also relates to a computer program comprising program code instructions for the execution of the steps of the method when the computer program is executed by a processing unit of a clock.

Other features and advantages will become apparent from the description given below, for informational purposes and without limitation in any way, with reference to the accompanying drawings.
1 shows a flow chart relating to a method for facial recognition of a wearer of a watch according to an embodiment of the present invention;
- Figure 2 shows a watch implementing this method according to an embodiment of the invention;

1 illustrates a method for facial recognition of a wearer of a watch 1 . This method aims in particular to perform reliable authentication of the wearer of the watch 1 on the basis of a three-dimensional modeling/representation of the wearer's face performed using the optical sensor 3 .

This method is implemented by the watch 1 , in particular a smart watch. This watch 1 illustrated in FIG. 2 is in a non-exclusive and non-limiting manner:

- processing unit (2);

- optical sensor (3);

- a hybrid display dial 4 provided with a first analog display component and a second digital and/or alphanumeric display component;

- a first light source 5 of light-emitting diode or laser diode type;

- a second light source 6 of the infrared type;

- a third light source (7) designed to be used as a photographic flash;

- Sound interface such as speaker;

- communication interface;

- behavioral or physiological movements; and

- contains behavioral and/or physiological sensors (8).

The processing unit 2 of the clock 1 comprises at least one processor cooperating with physical and software resources, in particular a memory element. These memory elements include a reference three-dimensional model for the wearer's face, information data for features characteristic of the wearer's face included in the reference three-dimensional model, and a reference identification index. This processing unit 2 is connected inter alia to the optical sensor 3 , the display dial 4 , the first, second and third light sources 5 , 6 , 7 , the sound interface, the communication module, and connected/connected to movement and to environmental, behavioral and/or physiological sensors 8 .

In this field of view 1 , the optical sensor 3 called a matrix image sensor may include sensors having a technology called CCD technology (charge transfer technology) or CMOS, for example. Such a sensor comprises a matrix of photosensitive image points (or pixels). This matrix of points is also referred to as the "pixel matrix" and contains NxM unit pixels which have very good sensitivity to light. In one alternative, such a sensor 3 may comprise a light-sensitive organic film instead of a conventional photosite. This optical sensor 3 also includes an analog-to-digital converter. This optical sensor 3 can implement an "auto-focusing function" and also an automatic "zoom" function as well as a function of capturing infrared images. In addition, the optical sensor 3 may operate according to the principle of time of flight (TOF) in order to measure a three-dimensional scene such as a wearer's face in real time, in this case, the optical sensor 3 may include an infrared band-pass filter to remove unnecessary ambient light.

Furthermore, the behavioral sensors 8 can measure all types of behavioral characteristics of the wearer of the watch 1 , such as, for example, movements or gestures performed by the wearer. To this end, these behavioral sensors 8 are one or more inertial sensors of the accelerometer, gyroscope or miniature multiple, capable of detecting angular velocity and linear accelerations along a plurality of axes associating the accelerometers and/or gyroscopes. It may include an axial gyrometer type, such as multi-axis sensors manufactured using MEMS technology. With respect to the physiological sensors 8 , these are functionalities of the wearer's tissue, such as, for example, pulse, skin impedance, blood pressure, respiration rate, respiratory arrhythmias, skin temperature, sweat rate, infrared radiation of the body, oxygen saturation of the blood or blood flow. parameters can be measured. The watch 1 may also comprise an environmental sensor specially configured to measure an environmental parameter, such as, for example, temperature, atmospheric pressure, or the like.

This method comprises a step (10) of initiation of an authentication process comprising a substep (11) of detection of at least one triggering movement/gesture performed by the wearer. This said at least one triggering movement/gesture is carried out by the part of the wearer's body comprising the watch during positioning of the wearer's face opposite the optical sensor 3 of this watch 1 . This step (10) aims to initiate the authentication process under certain conditions satisfied by the implementation of the substeps (11, 12) of the detection and determination of this step (10) described below.

The sub-step 11 of detection makes it possible to identify in advance the wearer of the watch 1 by ensuring that the wearer is, in fact, responsible for the authentication process that is about to begin. To this end, when the wearer wishes to be authenticated, the wearer executes at least one predefined movement/gesture, for example of the wearer's arm and/or the wearer's wrist provided with the watch 1 . This at least one movement/gesture is then performed by the wearer on at least one piece of measurement data originating from a behavioral sensor that captured this at least one movement/gesture of processing performed by the wearer. It is identified by the processing unit 2 after several operations have been performed. These operations include in particular comparing the at least one measurement data to one or more reference thresholds.

As mentioned above, this step 10 also includes a substep 12 of determining that the wearer's face located opposite the optical sensor 3 belongs to the living body. This sub-step 12 aims to efficiently verify that the face opposite the sensor is the face of the watch wearer who wants authentication, rather than a reproduction of the face that may be included in a photograph, video or photograph, for example. During this substep 12 , the processing unit 2 performs at least one physiological function performed on the face located opposite the optical sensor 3 for comparing the at least one piece of measurement data to one or more reference thresholds. Perform various processing operations on pieces of measurement data. This measurement is based on the emission of infrared radiation originating from this face, a change in the size of the pupil in at least one eye of this face after exposure to a third light source 6 serving as a photographic flash, the detection of blinking of the eyelids. can be established by

Note that this substep 12 of determining may be implemented simultaneously or substantially simultaneously with the substep 11 of detection.

If the initiating step 10 has been performed, the method then performs a capturing step 13 of at least one sequence of images of the wearer's face pivoting from one direction to another in front of the optical sensor 3 . includes doing More specifically, this step 13 is a substep 14 of acquiring a plurality of images of all parts forming the wearer's face when this face pivots from one direction to another in the front of the optical sensor 3 . ) is included. It is understood that the pivoting of the wearer's face in front of the optical sensor 3 is performed in such a way that all parts forming the face are captured. This substep 14 comprises a capturing step 15 of capturing at least two images for each same part of the face having two different light exposures, one exposure to visible light and one exposure to infrared light. It's about light.

In this condition, this substep 14 causes to obtain at least two sequences of images of the wearer's face, the surface of the first sequence of images of which is illuminated by visible light, and the second sequence of images of the face its surface is illuminated by infrared light.

The use of images comprising a face exposed to visible light facilitates the detection of features characteristic of the face relative to the skin color or texture of the face in these images. The use of images comprising a face exposed to infrared light is preferred because it is less dependent on lighting conditions than the use of images of the face captured in visible light.

These sub-steps (14) are:

- a watch in a stable or substantially stable position (1); and/or

- the wearer's face in an optimal position with respect to the optical sensor 3 in order to be included in each image of at least one sequence of images

visual and/or sound guiding (16) to the wearer of the watch (1) with the aim of maintaining

During this substep 14 , the processing unit 2 controls the focusing and “zooming” functions of the optical sensor 3 in order to contribute to the maintenance of the wearer's face in each image of the respective captured sequence of images. can be controlled

Thereafter, the method comprises a step of acquiring (17) surface geometric data of the face associated with each image of the at least one sequence while implementing the step of capturing at least one sequence of images of the wearer's face. During this step 17 , when the wearer's face pivots in front of the optical sensor 3 from one direction to another, this geometric data is generated for each acquired image comprised in the first and second sequences. . More specifically, during pivoting of the wearer's face, a beam of structured light of laser type is thus present in each image of the first and second sequences in such a way as to scan the three-dimensional shape of each of these parts of the face. It is projected onto each captured part of the face.

This acquisition step 17 is performed simultaneously or substantially simultaneously with the capture step 13 . Thereafter, the images of the first and second sequences are archived in memory elements of the processing unit 2 together with the corresponding geometric data associated with each of the images.

Thereafter, the method comprises a step 18 of generating ( 18 ) a three-dimensional model of the wearer's face from the obtained geometric data and from the sequence of at least one captured images. During this step 18 , the processing unit 2 implements an algorithm for construction of a three-dimensional model of the wearer's face on the basis of geometric data and at least one sequence of images. Such an algorithm uses the first and second sequences to correct defects that they may contain, such as peaks, troughs, or the presence of unwanted parts, such as hair or clothing, or accessories present on the face. It may include a preprocessing step of the images. Thus, this facet may include removal of peaks, filling of troughs, and correction of defects by segmentation of useful parts of the face. To do this, we can apply a 2D median filter to each image at the input to remove the noise. This noise may be in the form of pixels that are not coherent with their neighbors. This step can then involve the application of two-dimensional interpolation to fill the throats resulting from absorption of structured light by very simple opening of the mouth or by dark areas of the face such as the iris or eyebrows. On the basis of the images obtained for each of these sequences and the corresponding geometric data, a three-dimensional model is generated in particular by implementation of a three-dimensional meshing process.

The method then comprises a determining step 19 of an identification index generated on the basis of the identification data for a plurality of features characteristic of the face of the wearer of the watch 1 detected on the basis of the three-dimensional model . It enables three-dimensional analysis of the facial structure of features that are features of the face, also called point features of the face. That is, these characteristic features make it possible to interpret the face as a set of features of the feature dispatched by height, width and depth. Features of these characteristics are:

- about the geometry of the face;

- on the geometry of parts of the surface of the face;

- for the relative positions of parts of the face;

- about the texture of the skin or areas of this skin of the face;

- about the tone/color of these areas of skin on the skin or on the face

It may be in a non-limiting and non-exclusive manner.

For example, features of these features include wrinkles, chin, freckles, scars, lips, corners of lips, nose, tip of nose, ear, tip of ear, distance between eyes, distance between ridges of nose, distance between ridges , the distance between the jaws, and the like.

Preferably, features of this characteristic of the face that are taken into account for the determination of the identification index are comprised in the region/region of the wearer's face that remains stable during the representation of the face.

During this step 19 , a processing step 20 is implemented, during which identification data originating from the three-dimensional model is pre-written in the memory elements of the unit 2 to determine correspondence between these identification data. compared with the identified identification data. According to the result of this comparison, an identification index is generated by the processing unit 2 . This sub-step 20 of processing may involve the use of a limited configurable number of pieces of identification data for this implementation. This number can be parameterized according to the robustness desired to provide this method compared to the speed of performing this authentication method.

Accordingly, the method comprises a step 21 of identifying the wearer if the identification index is greater than the reference identification index. This reference index is located between the identification data of the three-dimensional model of the wearer's face constructed during the generating step 18 and the identification data pre-recorded in memory elements of the processing unit 2 relative to the reference three-dimensional model of the wearer's face. to verify the responsiveness of In other words, when the determined identification index is larger than the reference identification index, the generated 3D model has a significant number of similarities to the reference 3D model of the wearer's face. Conversely, when the determined identification index is equal to or less than the reference identification index, the number of similarities between the generated 3D model and the reference 3D model is not sufficient, and as a result, the wearer is not identified. In this case, a visual and/or sound message may be broadcast to the wearer informing the wearer that the wearer has not been identified.

The invention also relates to a computer program comprising program code instructions for the execution of steps 10 to 21 of the method when the computer program is executed by the processing unit (2) of the clock (1).

Claims (10)

A method for facial recognition of a wearer of a watch (1), comprising:
The following steps:
- a substep of detection (11) of at least one triggering movement/gesture performed by the wearer and a substep of (12) of determining that the face of the wearer located opposite the optical sensor (3) belongs to the living body; initiation of an authentication process comprising (10);
- capture (13) of at least one sequence of images of the wearer's face pivoting from one direction to another in front of the optical sensor (3);
- acquiring (17) surface geometric data of the face associated with each image of said at least one sequence;
- generating ( 18 ) a three-dimensional model of the wearer's face from the captured at least one sequence of images and from the surface geometric data obtained;
- determination (19) of an identification index generated on the basis of identification data for a plurality of features characteristic of the face of the wearer of the watch (1) detected on the basis of the three-dimensional model; and
- if the identification index is greater than the reference identification index, comprising an identification (21) of the wearer,
Physiological measurements for substep (12) of determining that the wearer's face belongs to the living body include: (i) emission of infrared radiation originating into the wearer's face, (ii) a third light source serving as a photographic flash (6) a change in the size of the pupil in at least one eye of the wearer's face after exposure to way for. delete The method of claim 1,
The step of capturing ( 13 ) is a sub of acquiring ( 14 ) a plurality of images of all parts forming the face of the wearer when this face pivots from one direction to another in front of the optical sensor ( 3 ). A method for facial recognition of a wearer of a watch (1), comprising the steps of: 4. The method of claim 3,
Said substep of acquisition 14 comprises a step of respective capture 15 of at least two images for each identical part of the face having two different light exposures, one exposure for visible light and one wherein the exposure of is to infrared light. 4. The method of claim 3,
The sub-steps of the acquisition (14) are:
- said watch (1) in a stable position; and/or
- the wearer's face in an optimal position with respect to the optical sensor 3 so that the face is included in each image of the at least one sequence of images
A method for facial recognition of a wearer of a watch (1) comprising the step of visual and/or sound guiding (16) of said wearer of said watch (1) to aim at maintaining The method of claim 1,
The method for facial recognition of a wearer of a watch (1), wherein the step of acquiring (17) is performed concurrently with the step of capturing (13). 7. The method of claim 6,
The step of determining (19) is a substep of processing (20) wherein the identification data derived from the three-dimensional model is compared with identification data pre-recorded against a reference three-dimensional model to determine correspondence between these identification data. A method for facial recognition of a wearer of a watch (1) comprising: 8. The method of claim 7,
said at least one triggering movement/gesture performed by a part of the wearer's body comprising said watch (1) during positioning of said wearer's face opposite said optical sensor (3) of said watch (1) A method for facial recognition of a wearer of a watch (1). 9. A watch (1) embodying the method according to any one of claims 1 and 3 to 8, comprising:
a processing unit (2), an optical sensor (3), at least one light source (5, 6, 7) and behavioral and/or physiological sensors (8), said processing unit (2) comprising said optical sensor ( 3) a watch (1) connected to the at least one light source (5, 6, 7) and to the behavioral and/or physiological sensors (8). A computer program stored on a computer-readable storage medium, comprising:
The computer program comprises program code instructions for the execution of the method according to any one of claims 1 and 3 to 8 when the computer program is executed by a processing unit (2) of a clock (1). , a computer program stored on a computer-readable storage medium.

Images